Asking ChatGPT about your own brand proves nothing. Logged-in memory, one-off prompts, and shortlist shuffles that change run to run produce screenshots, not data.
How to measure whether AI recommends you needs a fixed question set, clean sessions, four metrics, and a scoring rule you wrote before you looked at the results. Here is how to run the baseline, choose spreadsheet versus SaaS versus self-host, and report what boards actually need.